New-onset breast tenderness after initiation of estrogen plus progestin therapy and breast cancer risk

Abstract

Background

The use of estrogen plus progestin therapy increases both breast cancer incidence and breast tenderness. Whether breast tenderness during estrogen plus progestin therapy is associated with breast cancer risk is uncertain.

Methods

We analyzed data from the Women’s Health Initiative Estrogen plus Progestin Clinical Trial, which randomized postmenopausal women with an intact uterus to conjugated equine estrogens 0.625 mg plus medroxyprogesterone acetate 2.5 mg daily (CEE + MPA, N=8506) or placebo (N=8102). At baseline and annually, participants underwent mammography and clinical breast exam. Self-report of breast tenderness was assessed at baseline and 12 months. Invasive breast cancer incidence was confirmed by medical record review (mean 5.6 years of follow-up).

Results

Among women without baseline breast tenderness (N=14538), significantly more women assigned to CEE + MPA than placebo experienced new-onset breast tenderness after 12 months (36.1% vs. 11.8%, P<0.001). Among women in the CEE + MPA group, breast cancer risk was significantly higher in those with new-onset breast tenderness compared to those without new-onset breast tenderness (Hazard Ratio 1.48, 95% confidence interval 1.08–2.03, P=0.02). Among women in the placebo group, breast cancer risk was not significantly associated with new-onset breast tenderness.

Conclusions

New-onset breast tenderness during use of CEE + MPA was associated with increased breast cancer risk. The sensitivity and specificity for the association between breast tenderness and breast cancer were similar in magnitude to those of the Gail model.

Introduction

The use of menopausal estrogen plus progestin therapy is associated with increased breast cancer risk. In the randomized Women’s Health Initiative-Estrogen + Progestin Trial (WHI E+P trial), which compared daily use of oral conjugated equine estrogens 0.625 mg plus medroxyprogesterone acetate 2.5 mg (CEE + MPA) versus placebo, 8 additional cases of breast cancer/10,000 women-years occurred in women assigned to CEE + MPA compared to women assigned to placebo (P<0.003) ^{1, 2}. Identification of factors prognostic of breast cancer development in CEE + MPA users would have clinical relevance.

We hypothesized that breast cancer in this setting is associated with preceding breast tenderness. Breast tenderness is a common adverse effect of CEE + MPA. Although estimates of its frequency vary due to differences in data collection techniques, the preponderance of studies report that new-onset breast tenderness occurs in 10% to 25% of women following initiation of combined hormone therapy ^3–5. In addition, breast tenderness during combination hormone therapy use has been linked to an increase in mammographic density ^{3, 6, 7}, an independent risk factor for breast cancer ⁸.

The association between new-onset of breast tenderness during use of combined hormone therapy and breast cancer risk has not been previously examined. We examined this issue in the WHI E+P Trial, which randomized postmenopausal women to daily CEE + MPA or placebo ¹.

Methods

Study Design

Eligibility criteria and recruitment methods were previously described ^{1, 9}. For the WHI E+P Trial, 16608 postmenopausal women aged between 50 and 79 years without prior hysterectomy were recruited at 40 clinical centers largely by mass mailings between 1993 and 1998. WHI used the following criteria to define postmenopausal: no vaginal bleeding for 6 months (12 months for 50- to 54 year olds), hysterectomy, or past use of postmenopausal hormones ¹. Women were required to cease any menopausal hormone therapy for 3 months prior to randomization. Prior to enrollment, all women had a clinical breast exam and mammography; abnormal findings required clearance prior to study entry ². Interested WHI E+P participants were also enrolled in the WHI Dietary Modification component that randomly assigned 48835 postmenopausal women to a low-fat eating pattern or to usual diet ¹⁰. All participants provided written informed consent. Human subjects committees at each institution approved the study.

For the WHI E+P Trial, women were randomly assigned to a daily tablet containing CEE 0.625 mg and MPA 2.5 mg (N = 8506) or an identical-appearing placebo (N=8102). Local dispensation of study medications was blinded via medication bottles with unique bar codes. The following conditions required discontinuation of study medication: breast cancer, endometrial pathology (hyperplasia not responsive to therapy, atypia, or cancer), deep vein thrombosis, pulmonary embolism, malignant melanoma, meningioma, triglyceride level > 1000 mg/dL, institution of anticoagulant medication for thrombophlebitis, or use of nonstudy hormones (estrogen, progestin, androgen, tamoxifen, raloxifene), although short-term (<3 months) vaginal estrogen use was allowed.

Safety monitoring and assessment of adherence to therapy occurred 6 weeks after initiation of therapy and along with assessment of clinical outcomes at 6 month intervals. Annual mammography and clinical breast exam were required for continued administration of study medication. Participants were clinically monitored regardless of medication adherence.

After a mean of 5.6 years of follow-up, the WHI data and safety monitoring board recommended stopping the WHI E+P trial because breast cancer incidence exceeded a pre-designated stopping boundary and a global index supported a finding that overall risks exceeded overall benefits ¹.

Assessment of invasive breast cancer

Breast cancer outcomes were self-reported using standardized questionnaires provided every 6 months; breast cancer diagnoses were confirmed by local physician adjudicators who reviewed medical records and pathology reports. Subsequently, all breast cancer diagnoses were centrally-adjudicated by trained coders using standards from the Surveillance, Epidemiology, and End Results system ^{2, 11}. Community physicians evaluated and treated breast abnormalities. At the time study intervention ended, 359 invasive breast cancers were confirmed by central adjudication.

Assessment of breast tenderness

Breast tenderness was determined by self-report at baseline and at the 12-month follow-up visit through a symptoms inventory. The symptom checklist was based on questionnaire items related to menopausal hormone use and aging from national surveys and clinical trials ^{12, 13}. Participants rated the degree of bother from breast tenderness during the past 4 weeks using a four-point Likert-type scale: symptom did not occur, symptom was mild (did not interfere with usual activities), symptom was moderate (interfered somewhat with usual activities), symptom was severe (so bothersome that usual activities could not be pursued). We considered participants to have new-onset breast tenderness if they reported absence of breast tenderness at baseline and presence of breast tenderness (mild, moderate, or severe) at the first annual follow-up visit.

Other Questionnaire Measurements and Anthropometric Measures

At baseline, breast cancer risk factors were assessed by standardized self-report questionnaires. Participants were asked about medical and reproductive history, family medical history, cigarette smoking, alcohol use, race/ethnicity, education, income, and physical activity. Energy expenditure from recreational physical activity was calculated from questionnaire items regarding physical activity frequency and duration. Menopausal hormone use prior to trial intervention was ascertained at baseline with an interviewer-administered questionnaire. Gail breast cancer risk score was calculated based on risk factors collected at baseline ^{14, 15}. Baseline height and weight were measured for calculation of body mass index (BMI) i.e. weight (kg)/height (m²) (http://whiscience.org/about/collection.php).

Statistical Methods

All primary analyses were based on the intention-to-treat principle. Baseline characteristics were compared among women with and without baseline breast tenderness using Chi-squared tests of association. Statistical significance tests for baseline characteristics by new-onset of breast tenderness were adjusted for age and treatment assignment.

Relative risk of breast tenderness at first annual follow-up was obtained from a generalized linear model with a log-link function. We assessed whether the occurrence of breast tenderness at annual follow-up differed according to the presence of breast tenderness at baseline. To do this, we used a generalized linear model with randomization group (CEE + MPA vs. placebo) and baseline breast tenderness (yes/no) as main effects, risk of breast tenderness at first annual follow-up (yes/no) as the outcome, and a randomization group*baseline breast tenderness interaction term.

To examine the association between new-onset breast tenderness and invasive breast cancer risk in multivariate Cox models, we defined a time-dependent binary covariate X(t) equal to zero for all women until the 1^st annual follow-up, and equal to one after the 1^st annual follow-up if a woman reported breast tenderness at 1^st annual follow-up who had not reported breast tenderness at baseline. The survival time, t, was defined as the number of days after randomization to first diagnosis of breast cancer and censored at the time of a woman’s last documented follow-up contact or death. To allow a finer control for age, in addition to making a linear adjustment for age, we allowed the baseline hazard functions to vary by age group. Thus, each model allowed the baseline hazard function to vary by age group (50–54, 55–59, 60–69, or 70–79 years-old) and WHI diet modification trial randomization assignment and adjusted for age (linear), ethnicity (Caucasian, Black, American Indian, Asian Pacific Islander, unknown), alcohol consumption (non-drinker, ≤ 1 drink/day, > 1 drink/day), cigarette smoking (never, past, current), body mass index (kg/m², linear and quartiles), energy expenditure from physical activity (MET hours/week including walking, mold, moderate, and strenuous physical activity, linear and quartiles), parity (never pregnant, 1, 2, 3+), years of age at first birth (never pregnant, <20, 20–29, 30+), duration of breastfeeding (never, ≤ 1 year, > 1 year), years since menopause (< 5, 5–10, 10–15, >15), Gail model breast cancer risk (linear and quartiles), menopausal hormone therapy use prior to trial participation (yes/no), and baseline breast tenderness. Potential confounders were chosen based on biological plausibility and published studies ^16–21. Multiple imputation was used to avoid deletion of observations with missing covariate values ²². Hazard ratio estimates are expressed as relative risks. SAS PROC MI was used to generate a set of five plausible values for the missing covariate data that we assumed was missing at random and PROC MIANALYZE was used to combine parameter estimates from the Cox models for valid statistical inference.

To assess the extent to which new-onset breast tenderness was a marker of the effect of CEE + MPA on risk of invasive breast cancer, we compared the estimates of the CEE + MPA hazard ratio in models with and without new-onset breast tenderness ²³.

Using the method of Li and Meredith, we estimated the proportion of the effect of CEE + MPA on breast cancer risk that was explained by new-onset breast tenderness ²⁴. Using the incidence estimates of new-onset breast tenderness and the parameter estimates from a Cox proportional hazards model, this formula calculated the proportion of the treatment effect explained by the surrogate marker, new-onset breast tenderness, divided by the overall effect of treatment. We calculated the sensitivity (true-positive rate) and specificity (1 minus false-positive rate) of new-onset breast tenderness as a predictor of breast cancer risk among women assigned to CEE + MPA therapy.

All statistical tests were two-sided. The P-value threshold for statistical significance was 0.05 for tests of main effects. All statistical analyses were performed using SAS/STAT software Version 9.1 (SAS Institute, Inc, Cary, NC).

Results

For the 16,608 women participating in this trial, demographics and breast cancer risk factors (including prior hormonal exposure, family history, dietary intake, education, ethnicity, and Gail risk assessment) were balanced between the hormone therapy and placebo groups ².

Characteristics of women with and without baseline breast tenderness are compared in column P1 of Table 1. Characteristics of women with and without new-onset of breast tenderness are compared in column P2 of Table 1. Women reporting baseline breast tenderness tended to be younger, heavier, more often Black or Hispanic, lower alcohol consumers, less physically active, younger at first birth, more distant from menopause transition, at lower predicted breast cancer risk (Gail model), and more likely to have used menopausal hormone therapy prior to trial participation than women without baseline breast tenderness (Table 1). Baseline breast tenderness did not differ significantly by randomization group (Table 1).

Table 1.

Baseline Characteristics in the WHI E+P Trial by Breast Tenderness at Baseline and Year 1 (N=16608)¹

	Breast Tenderness (Baseline Status/Year 1 Status)
	No/No (N=10176)		No/Yes (N=3247)		Yes/No (N=717)		Yes/Yes (N=1037)
Characteristic	N	%	N	%	N	%	N	%	P1	P2
Age at screening, mean years (SD)	63.3	(7.1)	64.1	(6.9)	61.8	(7.2)	62.3	(7.1)	<0.001	<0.001
Treatment assignment
Placebo	5785	56.8	770	23.7	411	57.3	448	43.2	0.91	<0.001
CEE + MPA	4391	43.2	2477	76.3	306	42.7	589	56.8
Ethnicity
White	8840	86.9	2747	84.6	568	79.2	806	77.7	<0.001	0.001
Black	569	5.6	207	6.4	66	9.2	108	10.4
Hispanic	383	3.8	156	4.8	54	7.5	85	8.2
American Indian	27	0.3	10	0.3	4	0.6	8	0.8
Asian/Pacific Islander	229	2.3	75	2.3	11	1.5	21	2.0
Unknown	128	1.3	52	1.6	14	2.0	9	0.9
Alcohol intake
Non drinker	4186	41.3	1366	42.2	321	44.9	518	50.0	<0.001	0.25
≤ 1 drink/day	4560	45.0	1461	45.1	333	46.6	439	42.4
> 1 drink/day	1391	13.7	409	12.6	61	8.5	78	7.5
Smoking
Never smoked	5023	49.9	1636	50.9	351	49.6	518	50.4	0.88	0.03
Past smoker	3980	39.5	1306	40.6	284	40.2	400	38.9
Current smoker	1061	10.5	275	8.5	72	10.2	109	10.6
Body mass index (BMI), kg/m2 (collapsed categories)
<25	3203	31.7	1060	32.8	143	20.0	274	26.6	<0.001	0.04
25-<30	3528	34.9	1170	36.2	250	35.0	359	34.8
≤30	3386	33.5	1005	31.1	321	45.0	399	38.7
Quartiles of physical activity, MET hrs/week
< 1.5	2088	22.2	666	22.3	168	25.7	248	26.2	<0.001	0.47
1.5 ≥ to < 6.5	2181	23.2	734	24.6	164	25.1	250	26.5
6.5 ≥ to < 15.75	2461	26.2	740	24.8	156	23.9	238	25.2
≤ 15.75	2675	28.4	845	28.3	165	25.3	209	22.1
Parity
Never pregnant/Never had term pregnancy	1062	10.5	292	9.0	65	9.1	112	10.8	0.34	0.09
1	789	7.8	268	8.3	60	8.4	93	9.0
2	2217	21.9	693	21.4	154	21.5	248	24.0
3+	6064	59.8	1985	61.3	437	61.0	582	56.2
Age at first birth, years
Never pregnant/No term pregnancy	1062	11.5	292	9.9	65	10.4	112	12.1	<0.001	0.01
<20	1301	14.1	438	14.8	118	18.8	185	19.9
20 – 29	6015	65.0	1964	66.6	391	62.5	566	60.9
30+	871	9.4	256	8.7	52	8.3	66	7.1
Number of months breastfed
Never	4557	45.2	1439	44.9	331	46.6	513	49.9	0.001	0.85
≥ 1 year	3747	37.2	1190	37.1	270	38.0	379	36.8
> 1 year	1768	17.6	579	18.0	109	15.4	137	13.3
Years since menopause
<5	1564	16.7	398	13.3	153	23.8	187	19.8	<0.001	0.39
5 – 10 yrs	1799	19.2	540	18.1	134	20.9	213	22.5
10 – 15	1972	21.1	644	21.6	136	21.2	184	19.5
≤15 yrs	4023	43.0	1405	47.0	219	34.1	361	38.2
Quartiles of 5-year Gail model risk
< 1.1	1943	19.1	598	18.4	201	28.0	272	26.2	<0.001	0.05
1.1 ≥ 5 year risk< 1.44	2712	26.7	848	26.1	175	24.4	256	24.7
1.44 ≥ 5 year risk < 1.91	2818	27.7	881	27.1	149	20.8	239	23.0
5 year risk ≤ 1.91	2703	26.6	920	28.3	192	26.8	270	26.0
Prior hormone therapy use1
No	7572	74.4	2428	74.8	518	72.2	720	69.4	<0.001	0.73
Yes	2603	25.6	819	25.2	199	27.8	317	30.6

¹Of the 16608 WHI E+P trial participants, information regarding baseline and year 1 breast tenderness was available for 15177 participants. Thus, this table excludes participants for whom information regarding breast tenderness was missing from the baseline visit (N=95), the year 1 visit (N=1303), and both the baseline and year 1 visits (N=33).

^P1compares baseline characteristics of participants with versus without baseline breast tenderness (i.e., columns 1 and 2 vs. columns 3 and 4) based on chi-squared test of association for categorical variables and two sample t-tests for continuous variables.

^P2compares baseline characteristics in participants with versus without new-onset of breast tenderness (i.e., column 1 vs. column2) adjusted for age and treatment assignment. Tests of association for age and treatment assignment are unadjusted.

¹1045 (5%) of the 16,608 WHI E + P participants were current menopausal hormone therapy users at baseline, i.e. they underwent a hormone therapy “washout period” in order to participate in the WHI E + P trial.

As reported previously, assignment to CEE + MPA increased the risk of breast tenderness at 1-year follow-up ⁴. The risk of mild, moderate, or severe breast tenderness at 1 year in women assigned to CEE + MPA was higher among women without baseline breast tenderness than women with baseline breast tenderness (interaction term P-value < 0.001, Table 2). In the subgroup of women with no breast tenderness at baseline, the risk of breast tenderness at 1 year was 3-fold higher among women assigned to CEE + MPA compared to women who were assigned to placebo (36.1% versus 11.8%, relative risk [RR] 3.07, 95% confidence interval [95% CI] 2.85–3.30, P-value <0.001) (Table 2). In the subgroup of women with breast tenderness at baseline, the risk of breast tenderness at year 1 was 1.26 times higher among women assigned to CEE + MPA compared to women assigned to placebo (95% CI 1.17–1.37) (Table 2).

Table 2.

Prevalence and Relative Risk of Breast Tenderness at First Annual Follow-up by Randomization Assignment in the WHI E + P Trial (n=16608¹); all women and stratified by breast tenderness at baseline

	CEE+MPA		Placebo		RR2	(95% CI)	P-value3
	N	(%)4	N	(%)
All	3086/7808	(39.5)	1230/7464	(16.5)	2.40	(2.26, 2.54)	<0.001
							<0.001
No breast tenderness at baseline	2477/6868	(36.1)	770/6555	(11.8)	3.07	(2.85, 3.30)
Breast tenderness at baseline	589/895	(65.8)	448/859	(52.2)	1.26	(1.17, 1.37)

¹95 participants missing baseline breast tenderness data only, 1303 participants with missing year 1 data, and 33 participants missing both baseline and year 1 data.

²Relative risk of breast tenderness at 12-month follow-up from a generalized linear model.

³P-values for main effect of treatment (boldface) and for interaction between treatment assignment and baseline breast tenderness.

⁴Percent reporting breast tenderness at 12-month follow-up.

Women reporting new-onset of breast tenderness tended to be older and Black or Hispanic compared with women without new-onset of breast tenderness (Table 1, column P2). Most (76.3%) of the women who reported new-onset of breast tenderness were in the CEE + MPA treatment group. Of women in the CEE + MPA treatment group who reported new-onset breast tenderness, severity of tenderness was rated as mild in 77%, moderate in 19%, and severe in 4% (data not shown).

As reported previously, assignment to CEE + MPA increased the risk of breast cancer (HR 1.24, 95% CI 1.01–1.54, P=0.003) ². The new-onset of breast tenderness was associated with a higher risk of invasive breast cancer, especially among women who had been assigned to CEE + MPA (Table 3). Among women in the CEE + MPA group who did not have baseline breast tenderness, the risk of breast cancer was higher among women who had new-onset of breast tenderness than women without new-onset of breast tenderness (multivariable-adjusted HR 1.48, 95% CI 1.08–2.03, P-value = 0.02). Among women in the placebo group, the risk of breast cancer was not significantly higher among women who had new-onset of breast tenderness than among women without new-onset breast tenderness (Table 3). Thus, new-onset breast tenderness was associated with a statistically significantly elevated breast cancer risk in the CEE + MPA group, but not in the placebo group. Too few women had moderate or severe tenderness to allow us to determine whether breast cancer risk increased with increasing severity of breast tenderness.

Table 3.

Annualized rates and multivariable adjusted risk of invasive breast cancer due to new onset of breast tenderness at 12-month follow-up in the WHI E+P Trial

	Annualized rate of breast cancer
	Participants with new-onset breast tenderness		Participants with no new-onset breast tenderness
Treatment group	Cases (number at risk)	(%)1	Cases (number at risk)	(%)	HR2	(95% CI)	P-value
All	83 (3231)	(0.56%)	209 (10149)	(0.34%)	1.29	(0.99, 1.70)	0.06
Placebo	15 (760)	(0.44%)	111 (5769)	(0.33%)	0.99	(0.59, 1.66)	0.97
CEE +MPA	68 (2471)	(0.60%)	98 (4380)	(0.36%)	1.48	(1.08, 2.03)	0.02

¹Annualized rates (unadjusted).

²Hazard ratio from Cox proportional hazards models comparing risk of breast cancer among women with versus without new-onset breast tenderness. Cox proportional hazards models are adjusted for CEE+MPA randomization assignment, age, ethnicity, alcohol consumption, smoking, body mass index (linear and quartiles), physical activity (linear and quartiles), parity, age at first birth, breast feeding, years of age at menopause, Gail model breast cancer risk (linear and quartiles), and menopausal hormone therapy use prior to trial participation. This table displays hazard ratios for women without baseline breast tenderness.

Figure 1 illustrates the difference in invasive breast cancer risk according to presence versus absence of new-onset of breast tenderness after adjustment for age, ethnicity, prior menopausal hormone therapy use, and Gail breast cancer risk score. The data were too sparse to allow for a stratification of the breast tenderness-breast cancer association according severity (mild, moderate, severe) of breast tenderness.

Figure 1. Risk of invasive breast cancer by new-onset of breast tenderness.

Survival times for Kaplan-Meier estimates of cumulative hazard function beginning at the year 1 follow-up visit (i.e. survival time t=0 at the year 1 follow-up visit). Adjusted for age, ethnicity, prior menopausal hormone therapy use, and Gail breast cancer risk score. Seventeen women in the active arm (6 with breast tenderness at year 1, 11 without breast tenderness at year 1) and 26 women in the placebo arm (10 with breast tenderness at year 1, 16 without breast tenderness at year 1) were excluded from the Figure because they either developed breast cancer prior to the year 1 follow-up assessment or had no follow-up after their annual visit.

We used Prentice’s criteria to further explore the breast tenderness-breast cancer association ²³. Prior to inclusion of randomization assignment, women with new-onset of breast tenderness had a 37% higher risk of invasive breast cancer than women without new-onset of breast tenderness (HR 1.37, 95% CI 1.05–1.77, P-value = 0.02, data not shown). When both treatment assignment and new-onset of breast tenderness were included in the same Cox regression model, the magnitude of the increased risk of breast cancer among women with new-onset of breast tenderness was modestly lowered (HR 1.29, 95% CI 0.99–1.70, data not shown). Likewise the magnitude of the HR associated with assignment to CEE + MPA was lowered and lost statistical significance after adjustment for new-onset of breast tenderness (HR 1.25 and P-value = 0.04 prior to adjustment, HR 1.19 and P-value = 0.11 after adjustment for new-onset of breast tenderness, data not shown).

Among women assigned to CEE + MPA therapy, new-onset breast tenderness had a sensitivity of 41%, a specificity of 64%, and a positive predictive value of 2.7% in predicting invasive breast cancer risk during the intervention period (mean 5.6 years). The proportion of breast cancer risk conferred by CEE + MPA that was explained by new-onset of breast tenderness was 24% (95% CI 6%-100%), as assessed by the method of Li and Meredith.

Discussion

In participants of a large randomized controlled trial of menopausal hormone therapy, new-onset of breast tenderness was a marker of future breast cancer risk. Among women assigned to the CEE + MPA group who were free of breast tenderness at baseline, those who reported new-onset of breast tenderness at first annual follow-up had a 48% higher risk of invasive breast cancer than did those who did not experience new-onset of breast tenderness at first annual follow-up. New-onset of breast tenderness explained 24% of breast cancer risk conferred by CEE + MPA.

To our knowledge, no prior published studies have addressed whether there is an association between CEE + MPA-induced new-onset breast tenderness and breast cancer risk. However, an association between new-onset breast tenderness and breast cancer has biological plausibility. Three prior studies found that CEE + MPA -induced breast tenderness is associated with increased mammographic density ^{3, 6, 7}, a risk factor for breast cancer which indirectly measures breast parenchymal tissue proliferation ^25–27. Studies have linked administration of estrogen + progestin therapy with increased breast cell proliferation ^28–30. Thus, breast discomfort may be a clinical manifestation of increased proliferation that is manifest radiographically as increased breast density. However, in one study, mammographic density did not mediate the association between combination hormone therapy and breast cancer ³¹. The present study design does not permit us to directly test whether combined hormone therapy-induced breast tenderness represents increased breast cell proliferation. Because increases in serum level of estrone and estrone sulfate during CEE + MPA therapy are positively associated with increases in mammographic density ^{32, 33}, it is possible that CEE + MPA-induced increases in serum estrone or estrone sulfate level could result in both breast tenderness and increased breast cancer risk.

Emergence of endocrine symptoms in response to hormone-based interventions has been recently linked to breast cancer outcome in a large adjuvant breast cancer trial. In the Anastrazole, Tamoxifen, Alone and Combined (ATAC) trial, breast cancer patients reporting an increase in hot flashes and arthralgia symptoms after 3 months of therapy with tamoxifen or aromatase inhibitor had nearly half as many breast cancer recurrences compared to women not reporting such symptoms ³⁴. Thus, systematic responses such as the breast tenderness described in the current report may represent integrated functions of biological interaction among the intervention, breast cancer, and the postmenopausal woman’s host response.

In our study, the prevalence of new-onset breast tenderness among women assigned to CEE + MPA was 36%. Most randomized trials of CEE 0.625 mg/d + MPA 2.5 mg/d did not report the prevalence of new-onset breast tenderness at 1 year ^35–40. The exception is the Postmenopausal Estrogen/Progestin Interventions Trial, PEPI, which asked participants at the 1^st annual follow-up: “During the past week including today, did any of these symptoms bother you or interfere with your life? Breast sensitivity/tenderness (yes or no) and/or painful breasts (yes or no)”. In PEPI, among women assigned to CEE + MPA who reported absence of baseline breast symptoms, the prevalence of breast symptoms was 23% at 1^st annual follow-up ³. When we exclude WHI participants with mild breast tenderness, more closely matching the PEPI assessment method, the prevalence of hormone therapy-associated new-onset breast tenderness is virtually identical in the two trials (23% vs. 25%). Thus, it is likely that heterogeneity regarding how questionnaires assess breast symptoms results in different prevalence estimates for hormone therapy-associated breast symptoms across studies.

Few characteristics explained who would develop new-onset breast tenderness. Although several characteristics were statistically significantly associated with new-onset breast tenderness (e.g. ethnicity, cigarette smoking, body mass index), the magnitude of their associations with breast tenderness was small. Risk factors for new-onset breast tenderness require further elucidation.

The sensitivity and specificity of new-onset breast tenderness for predicting invasive breast cancer risk among CEE +MPA users were similar to that of the Gail model. In our study, based on a mean follow-up duration of 5.6 years, new-onset breast tenderness had a sensitivity of 41%, a specificity of 64%, and a positive predictive value of 2.7% for predicting invasive breast cancer risk among women assigned to CEE + MPA. Using a threshold of 1.67% five-year risk of developing invasive breast cancer, the Gail model had a sensitivity of 44% and a specificity of 66% ⁴¹. In another study, a positive mammogram had a positive predictive value of 6.6% for women aged 50 to 59, and 7.8% for women aged 60 to 69 years ⁴².

Our findings have potential clinical implications. The WHI E+P Trial has previously demonstrated that combined hormone therapy increased invasive breast cancer risk ^{2, 43}, increased mammographic breast density ⁴⁴, and increased the frequency of mammograms with abnormalities which less reliably detect cancer ⁴⁵. We now report that an increase in breast tenderness, easily detected by physicians or patients, identifies a population at particular risk of breast cancer development. These findings should be considered by women who experience new-onset breast tenderness during combined hormone therapy use and their prescribing physicians to inform decisions regarding continued combined hormone therapy use.

Our study has limitations. The study questionnaire assessed breast tenderness annually. Thus, we may have underestimated breast tenderness, although this method of ascertainment probably resembles reporting of breast tenderness in a clinical setting. Also, although the rates of treatment discontinuation in the CEE + MPA arm (42%) and of cross-over from placebo to active therapy (11%) were relatively high ¹, we believe that they would tend to decrease the observed association between breast tenderness and breast cancer. Although women who developed new-onset breast tenderness had higher baseline Gail model breast cancer risk scores, the association between new-onset breast tenderness and future breast cancer risk persisted after adjustment for Gail risk score. Finally, our results do not apply to other types or schedules of estrogen or progestogen therapy.

Strengths of our study include the large number of participants, the use of placebo controls, the comprehensive assessment of breast cancer risk factors, the rigorous assessment of breast cancer outcomes over several years of follow-up, the blinding of participants and investigators to treatment assignment, the requirement for annual mammography and clinical breast exam, and the serial prospective blinded assessment of breast tenderness in both placebo and treatment groups. This study was based on the largest and longest randomized controlled trial of combination menopausal hormone therapy ever performed.

In conclusion, the new-onset of breast tenderness during CEE + MPA may be a marker of increased breast cancer risk. The sensitivity and specificity for the association between breast tenderness and breast cancer were similar in magnitude to those observed with the Gail model and raise the possibility that reports of breast tenderness during CEE + MPA therapy may identify high-risk women.